【摘要】：近年來,隨著信息技術(shù)的發(fā)展和普及,計(jì)算機(jī)網(wǎng)絡(luò)的拓?fù)湟?guī)模不斷增大,其上承載的業(yè)務(wù)應(yīng)用也在不斷擴(kuò)展,這樣使得網(wǎng)絡(luò)的異構(gòu)性和復(fù)雜性越來越高。特別是對于政府、軍隊(duì)、企業(yè)、院校所建立的大型三級局域網(wǎng)來說,這些現(xiàn)象更加地突出,迫切需要優(yōu)異的性能檢測系統(tǒng)。傳統(tǒng)的網(wǎng)絡(luò)性能檢測工具不僅需要路由器等網(wǎng)絡(luò)內(nèi)部節(jié)點(diǎn)的協(xié)作,而且無法描述網(wǎng)絡(luò)的整體性能。近年來,在現(xiàn)代醫(yī)學(xué)、地理領(lǐng)域比較成熟的斷層掃描技術(shù)開始被廣泛應(yīng)用于網(wǎng)絡(luò)性能檢測。斷層掃描技術(shù)的基本原理是利用網(wǎng)絡(luò)端到端的性能探測結(jié)果和探測路徑矩陣推斷出網(wǎng)絡(luò)的內(nèi)部鏈路性能。然而三級網(wǎng)的規(guī)模巨大,終端節(jié)點(diǎn)數(shù)目較多,直接在全網(wǎng)做端到端測量存在探測開銷大、計(jì)算復(fù)雜度高的問題。針對這些問題,本文在對傳統(tǒng)網(wǎng)絡(luò)測量技術(shù)和斷層掃描技術(shù)研究的基礎(chǔ)上,結(jié)合三級網(wǎng)拓?fù)浜蜆I(yè)務(wù)應(yīng)用特點(diǎn),設(shè)計(jì)了一種基于斷層掃描技術(shù)的三級網(wǎng)丟包檢測與分析系統(tǒng),并對其關(guān)鍵技術(shù)進(jìn)行了研究。首先,根據(jù)三級網(wǎng)的拓?fù)浣Y(jié)構(gòu)和業(yè)務(wù)應(yīng)用特點(diǎn)設(shè)計(jì)了一種三級網(wǎng)丟包檢測與分析系統(tǒng)。由于三級網(wǎng)的每個(gè)上層管理域包含多個(gè)下層管理域,系統(tǒng)可以利用各個(gè)管理域丟包檢測的并行性,使用分布式與集中式相結(jié)合的方法在不同級別層次上進(jìn)行協(xié)同的網(wǎng)絡(luò)丟包檢測與分析,然后逐級歸約分析結(jié)果,從而得到逐步擴(kuò)大和細(xì)化的網(wǎng)絡(luò)丟包視圖。其次,提出了基于最短鏈路序列的丟包率計(jì)算方法及最短鏈路序列的逐級歸約算法。由于探測路徑規(guī)模的限制,現(xiàn)有的網(wǎng)絡(luò)斷層掃描計(jì)算方程存在欠定性,不能求出鏈路級的精確丟包率,因此提出基于最短鏈路序列計(jì)算丟包率,并在計(jì)算結(jié)果的逐級合并歸約中逐步擴(kuò)大丟包檢測視圖,逐步細(xì)化丟包計(jì)算結(jié)果。實(shí)驗(yàn)數(shù)據(jù)表明該模型可以通過較小的開銷來檢測目標(biāo)網(wǎng)絡(luò)的丟包情況。再次,在最短鏈路序列不能滿足當(dāng)前丟包檢測的精度需求時(shí),提出了動態(tài)探測路徑集合構(gòu)造算法。該算法的主要思想是對待選探測路徑按信息增益進(jìn)行排序,每輪選擇固定數(shù)量的端到端探測路徑,并基于探測結(jié)果細(xì)化最短鏈路序列和相應(yīng)的丟包信息。當(dāng)待求解鏈路數(shù)量減少到一定規(guī)模的時(shí)候,直接啟動預(yù)先選擇構(gòu)造算法構(gòu)造探測路徑集合并發(fā)送探針到網(wǎng)絡(luò)中,進(jìn)而求解出所有鏈路的丟包率。實(shí)驗(yàn)數(shù)據(jù)表明該算法以較小的探測開銷在很短的時(shí)間內(nèi)細(xì)化鏈路序列的丟包信息,達(dá)到探測與分析的粒度要求。最后設(shè)計(jì)實(shí)驗(yàn)驗(yàn)證了所設(shè)計(jì)的系統(tǒng)和所提出的算法的有效性。
[Abstract]:In recent years, with the development and popularization of information technology, the topological scale of computer networks is increasing, and the business applications carried by them are also expanding. This makes the heterogeneity and complexity of networks more and more high. Traditional network performance detection tools need not only the cooperation of nodes in the network such as routers, but also can not describe the overall performance of the network. The basic principle of the technique is to deduce the network's internal link performance from the network's end-to-end performance detection results and the detection path matrix. However, the scale of the three-tier network is huge and the number of terminal nodes is large. Based on the research of network measurement technology and tomography technology, a three-level network packet loss detection and analysis system based on tomography technology is designed, and its key technologies are studied. Firstly, a three-level network is designed according to the topological structure and service application characteristics of the three-level network. Packet loss detection and analysis system. Since each upper management domain of the three-tier network contains multiple lower management domains, the system can use the parallelism of packet loss detection in each management domain to detect and analyze network packet loss in different levels by combining distributed and centralized methods, and then reduce the analysis results step by step. Secondly, a packet loss rate calculation method based on the shortest link sequence and a step-by-step reduction algorithm based on the shortest link sequence are proposed. The experimental data show that the model can detect packet loss in the target network with less overhead. Thirdly, the shortest link sequence can not meet the current packet loss detection. The main idea of this algorithm is to sort the selected paths according to information gain, select a fixed number of end-to-end detection paths per round, and refine the shortest link sequence and the corresponding packet loss information based on the detection results. The experimental data show that the algorithm can refine the packet loss information of the link sequence in a very short time with a small detection overhead to meet the granularity requirements of detection and analysis. The effectiveness of the proposed system and the proposed algorithm is verified by experiments.
【學(xué)位授予單位】：國防科學(xué)技術(shù)大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2014
【分類號】：TP393.06;TP393.1
，

本文編號：2183028